One of the main components in an extrusion line is the die, which has significant influence on the product quality. The wall thickness distribution of an extruded tube for example is directly influenced by the melt distribution in the die, so in the process of die design the resulting product quality is set. Nowadays three-dimensional simulations of the complex flow in the die by usage of computational fluid dynamics (cfd) are used in the design process. The results of such cfd-simulations have to be analyzed for every design proposal by experienced engineers. The geometry optimization only based on the basic post processing tools is very time consuming and depends on the background knowledge of the cfd-engineer. Existing approaches deal with the automation of this process. There are two main methods known, one is based on one dimensional analytical equations and the other one is based on three dimensional cfd-simulations. Both methods compete against each other in relation to the calculation time and accuracy of describing the real flow conditions. The content of this publication is the optimization of spiral mandrel dies by a coupled approach. An initial dimensioning based on a flow analysis network (fan) has been conducted. The result is an initial geometry of a spiral mandrel die for defined process conditions and general dimensions of the die. In addition to that a bandwidth is calculated in which an automatic optimization based on three dimensional cfd-simulations takes place. Therefor quality criteria are defined which quantify the die quality based on scalar values for the volume flow distribution, relation of wall shear stresses and the deviation of the spiral channel volume flow to a defined ideal flow. The optimization process is based on a multicriterial genetic optimization algorithm. The geometry of the spiral mandrel die is described by a three dimensional cad-model, so the manufacturability of the resulting optimum is saved.